Research On Container Throughput Forecasting Based On Nonlinear Ensemble Method And Error Correction

In the context of economic globalization,containerization plays a crucial role in global trade and competition among countries.With the development of container transportation,it also greatly promotes the sustainable and rapid development of port modernization.At the same time,there are a series of problems such as overcapacity,throughput utilization decline.Since the cost of port construction is sunk cost,over-construction will cause huge economic loss and resource waste.Accurate forecasting of container throughput is particularly important in port transport systems and operating policies.However,due to the complexity of factors affecting the variation of container throughput,container throughput data are characterized by nonlinearity,nonstationarity,high volatility,irregularity and randomness,etc.,so it is a complex and challenging process to model and predict the demand of cargo transportation at the port level.Therefore,how to construct a scientific and effective forecasting model to improve the forecasting accuracy of container throughput is of great theoretical and practical significance.In this paper,based on the existing theories and the characteristics of container throughput data,by combining with the decomposition-ensemble idea,BP neural network,the particle swarm optimization algorithm(PSO)and the grey wolf optimization algorithm(GWO)as the main technologies,put forward two new nonlinear ensemble forecasting methods respectively: container throughput forecasting method with nonlinear ensemble,container throughput forecasting method with decomposition-ensemble and error factor correction.The proposed two forecasting methods can effectively improve the prediction accuracy of container throughput and provide reference basis for port's operation system and operation decision in practice.This paper mainly discusses the following two parts:(1)Container throughput forecasting method with nonlinear ensemble.Based on the idea of decomposition-ensemble,a new nonlinear ensemble combination model is proposed.The method using the ensemble empirical mode decomposition(EEMD),BP neural network optimized by particle swarm optimization(PSO-BP)and BP neural network optimized by grey wolf optimizer(GWO-BP)is constructed based on EEMD-PSOBP-GWOBP nonlinear ensemble container throughput forecasting method,and the method is applied to forecast the container throughput in Shanghai,Shenzhen and Qingdao.The empirical results show that the proposed method significantly improves the level accuracy and direction accuracy of container throughput forecasting,which proves the importance of nonlinear ensemble in forecasting process.(2)Container throughput forecasting method with decomposition-ensemble and error correction.Based on the processing method of forecasting error in error correction,adecomposition-ensemble forecasting method with high complexity error factor correction(HEFC)pattern is proposed,which can extract the useful information of forecasting error effectively and reduce the superposition of system error.The method using variational mode decomposition(VMD),complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN),BP neural network optimized by hybrid particle swarm optimization and grey wolf optimization algorithm(PGBP)to build a model based on VMD-PGBP-HEFC decomposition-ensemble container throughput forecasting method.The proposed method is remarkable in Hong Kong,Shanghai and Singapore ports,which confirms the superiority of the HEFC pattern in reducing the system error.